Stay curious.

Listen before trying to persuade.

Being listened to well lowers defences. In a 2025 field experiment, people who were listened to well thought better of the person listening and felt less defensive, though they were no easier to persuade (Porat and colleagues 2025). Understanding someone and changing their mind are different outcomes.

Make room for uncertainty.

Intellectual humility is recognising the limits of what we know (Leary and colleagues 2017). It does not require abandoning a belief or finding a middle position.

Reflection should lead to insight.

More introspection is not always better (Grant and colleagues 2002). Notice whether reflection is helping, and give yourself permission to stop when it becomes repetitive or distressing.

Built for human agency.

No bots, no engagement feed, no public popularity scores and no hidden ranking of opinions. Emotions are information, not proof that a belief is wrong. You choose what to talk about and whether to meet someone who sees it differently or someone who sees it the same way.

Dialogue records a Dialogue only when everyone speaking says yes on their own screen, and writes one down only once they all keep it. Nobody listens in to a one-to-one Dialogue unless you both turn on “Let people listen” during the call, and either of you can turn it off again. Been there is never open to listeners. A hosted Dialogue is open to listeners from the start. A group Dialogue open to every member can be heard without joining unless its host kept it to the people in the Dialogue. Every Dialogue says who can hear it before you join.

Every scale we use is a public-domain measure from the International Personality Item Pool (IPIP), a research collection maintained at the Oregon Research Institute. The questions are short, plain and free for anyone to use. We chose them because they are open, well studied and quick enough to answer on a phone.

Before your first Dialogue you answer ten short questions taken from these scales: two each on sociability, speaking up, warmth and enjoying hard thinking, one on getting things done and one on imagination. After your first Dialogue you can answer the full set, 81 questions and about 9 minutes in all. It starts with sixteen questions from the Mini-IPIP (Donnellan and colleagues 2006) on four of the five broad personality traits, then seven short scales. The four marked “Helps pairing” describe how you take part in a conversation. The other three only shape How you talk.

• IPIP Need for Cognition, 10 questions, modelled on Cacioppo and Petty’s 1982 construct: how much you enjoy effortful thinking. Helps pairing.
• IPIP-VIA Judgment/Open-mindedness, 9 questions: weighing reasons before deciding. Helps pairing.
• IPIP Attending to Emotions (Barchard 2001), 10 questions: noticing your own feelings.
• IPIP-NEO Emotionality, 10 questions: feeling things strongly, including other people’s feelings.
• IPIP-NEO Assertiveness, 10 questions: taking charge and speaking up. Helps pairing.
• IPIP-HEXACO Flexibility (Ashton, Lee and Goldberg 2007), 10 questions: how easily you are convinced. Helps pairing.
• IPIP-HEXACO Patience, 10 questions: how quickly heat rises for you.

Some scales share a question with another, for example one statement about noticing your emotional reactions appears in both feelings scales. A shared statement is asked once and counts for both. The two feelings scales are averaged into one lean on purpose, the same way the two thinking scales are, because two short scales describe one idea more reliably than one.

We do not measure anxiety, loneliness, honesty, values or anything meant to sort people into better and worse. We no longer ask the Mini-IPIP’s four emotional stability questions. Three were about mood. The fourth, “Get upset easily”, is asked once, among the patience questions, where it shapes only the private note about heat.

Your answers add up to three leans that matter in a conversation. Speaks up early or Lets others go first comes from the questions about taking charge and speaking up, with the four about being outgoing from the Mini-IPIP. Feeling first or Reasons first sets the questions about noticing and feeling emotions against the ones about enjoying hard thinking and weighing things up. Both groups are shown on their own too, so describing neither strongly never reads as sitting in the middle. Takes a lot to move or Open to being moved comes from the questions about taking advice and being convinced.

Each lean is a line with a side at either end, and your answers put a dot on it. Around the middle of each line is a shaded zone as wide as the questions’ own margin of error. A dot inside it could land on the other side on a second try, which the page says. The line sits at a fixed point for now and will move to the middle of Dialogue’s own members once enough people have answered.

The sides of the three leans make eight personas, one for each combination. Finish the full set and yours appears under You, with how you tend to connect, what to watch for, the kinds of work people who talk alike are often drawn to and the research behind each line. You can also answer 71 of the questions without an account at How do you talk?, where everything stays in your own browser.

How sure we are

The questions behind each lean agree with each other well, though any short set of questions is rough. If people answered again a month later, most would stay on the same side of each line. The ones who crossed would mostly be the ones inside the shaded zone. A persona is a friendly way to describe the three sides together rather than a type. The lines drawn to other research are our reading of it, and each one says how sure we are.

A fourth note, how quickly heat rises, comes from the patience questions. Only you see it. It never changes your persona and never appears anywhere else.

After you have had a few Dialogues, what actually happens starts to count too: whether you changed your answer after talking with someone who saw it differently, whether partners say they felt heard and whether you take a listening or speaking seat in group Dialogues. Your answers keep at least 70 percent of the weight. If your Dialogues and your answers point different ways, we tell you, because that is usually the most interesting thing about a person.

How you talk stays private. It is never used for pairing, ranking or deciding what you can join.

Pairing uses short questions and how you like to talk. It only ever introduces two people who fit what each of them chose, in the same language. On a statement you answer Agree, Not sure or Disagree and choose who to meet: someone who sees it differently or someone who sees it the same way. Not sure counts as different from a settled answer. Two people who are both not sure see it the same way. A story meets a story on the same theme, a decision meets someone who has already made the same call and Been there meets someone who knows the same subject. After a short wait a statement can widen to another statement you have both answered, and a story to any story.

Among the people who fit, your answer counts most. If you asked for someone who sees it differently, someone with the opposite answer comes before someone who is not sure. How you like to talk also counts: the pace you prefer and what draws you in, both set under You. A small experimental part then looks at how your answers to the short questions fit together:

• similar airtime, from the sociability questions together with taking charge and speaking up;
• a similar thinking style, from the imagination questions together with enjoying hard thinking;
• similar conscientiousness;
• at least one of you open to other views, from weighing things up together with taking advice and being convinced;
• at least one of you high in agreeableness.

Airtime follows the finding that strangers who are alike in sociability get on a little better. Thinking style and conscientiousness apply the same idea, which is still unproven. The last two follow the finding that two people who are both disagreeable have the least pleasant conversations, so what matters is that at least one of you brings the warmth or the openness (Cuperman and Ickes 2009). The ten questions before your first Dialogue give only a rough first reading, and the longer scales take over as you answer them. Pairing does not use your persona, your emotional stability, how strongly you feel things, how much you notice your feelings or how calm you stay, so nobody is paired or ranked on temperament. A scale either of you has not answered is simply left out. We are still finding out whether any of this makes conversations better. We do not yet know, and we will say so until we do.

On similarity: in a study of first meetings between strangers, Cuperman and Ickes (2009) found that two extraverts or two introverts got on somewhat better than mixed pairs, while pairs in which both people were low in agreeableness had the least pleasant conversations. The effects were modest and the conversations were short. In large samples of couples, Dyrenforth and colleagues (2010) found that similarity in personality added almost nothing to satisfaction once each person’s own traits were counted. Similarity is not the magic ingredient people assume.

On disagreement: Minson, Chen and Tinsley (2020) show that people differ in how receptive they are to opposing views, and Yeomans and colleagues (2020) show that receptive language, acknowledging the other view, hedging and finding points of agreement, makes a person more persuasive and more worth continuing with. Short one-to-one conversations can shift attitudes a little, and the shift can last (Kalla and Broockman 2020). Political opponents who talked about their own lives warmed toward the other side for a while, though talking over their disagreement did not do the same (Santoro and Broockman 2022). These are small effects, produced by how people talk rather than by who they are.

On listening: Itzchakov, Kluger and Castro (2017) and Itzchakov and colleagues (2018) found that being listened to well makes speakers less defensive, clearer about their own view and more able to hold its contradictions.

On talking to strangers: Epley and Schroeder (2014), Sandstrom and Boothby (2021) and Kardas, Kumar and Epley (2022) all found the same thing. People expect conversations with strangers, and deeper ones especially, to be more awkward and less enjoyable than they turn out to be.

On prediction: Joel, Eastwick and Finkel (2017) fed a large speed-dating dataset to machine learning and could predict who tends to be liked in general, but not which two particular people would like each other. Joel and colleagues (2020) found across 43 couples studies that how people see their relationship predicts its quality far better than either partner’s personality. We take this seriously. Who you are helps us set the table. It does not predict the meal.

Your persona is not a validated type. Nobody has shown that these eight personas predict anything, and we will not claim they do. It says nothing about your health or your worth and it ranks nobody. Questionnaires like these are reliable enough to describe a lean, and rough enough that you may land on the other side of a line on another day, which is why each line shows a shaded zone and why your Dialogues are allowed to move the picture. The persona is a friendly way of describing how you tend to talk. What you do with it is entirely up to you.